Chemistry and Physics. 479 



lower end of this tube projects below tbe surface of the mercury 

 in a reservoir open to the outside air. Water vapor at about 

 atmospheric pressure flows down around the outside of the porce- 

 lain cylinder and escapes, for the most part, into the surrounding 

 atmosphere. The rest of the water vapor diffuses inward through 

 the walls of the porcelain cup, is condensed by the cooler, and 

 runs down into the collecting bulb. At the same time the air 

 imprisoned between the outside of the cooler and the inside of 

 the porcelain vessel diffuses outward in the direction opposite to 

 that of the diffusing water vapor. The whole apparatus acted as 

 a pump, and the column of mercury rose in the barometer tube 

 until it attained a height 15 mm less than the "barometric height." 

 This difference of 15 mm was due to the pressure of water vapor 

 corresponding to the temperature of the cooler. The action of 

 the pump may ba better understood from the following element- 

 ary considerations. The water vapor on the inside of the porce- 

 lain cylinder is prevented from exerting a pressure greater than 

 15 mm by being condensed and removed from the field of action 

 before it can accumulate. The air in the same vessel diffuses out- 

 wards as long as any of it remains. As soon, however, as the 

 air has passed through the porcelain walls it is swept out by the 

 current of water vapor and escapes into the outside atmosphere 

 before it can exert any appreciable back pressure. Finally all of 

 the air will have diffused out of the porcelain cup and the only 

 source of pressure will be the water vapor at about the tempera- 

 ture of the cooler. As Gaede says: "This experiment shows for 

 the first time that one can produce a continuous sucking action 

 by means of the diffusion of vapors and gases alone, without the 

 motion of solid or liquid pistons." The fatal fault of this particu- 

 lar form of pump is the very long time required to reach the 

 lowest pressure which the system is capable of producing. Gaede 

 shows, however, both theoretically and experimentally that holes 

 in (metallic) diaphragms will give the highest rate of evacuation 

 when the diameter of the opening is of the order of magnitude 

 of the mean free path of the molecules of the gas. He therefore 

 replaces the porcelain membrane by slits between metallic jaws. 

 Also, for obvious reasons, mercury vapor is used in the final form 

 of pump. An especially noteworthy result, — which is also pre- 

 dicted by the theory of diffusion, — follows from the tables of test 

 data given in the original article. It is that the speed of evacua- 

 tion remains constant even to the highest vacua obtainable. This 

 property has never before been observed in any other type of air 

 pump. A few typical numerical data illustrating the general 

 behavior and efficiency of the new diffusion pump will now be 

 given. Beginning with the pressure of 0'25 mm a Rontgen ray 

 bulb containing air was evacuated in 25 sec. to the extent that 

 the current preferred a parallel spark gap of 16 mm length. On 

 further pumping, the spark gap increased to 100"' m . In this case 

 the temperature of the mercury vapor was 119°C. The corre- 

 sponding times at temperatures of 126° and 135° were 30 and 45 



